Case Study A
How to read an Evidence Record
This case uses an in-corpus secondary-outcome record from NCT01441947. It is a compact mechanism demo: five changed paths, one outcome rule, one timing modifier, and explicit claim boundaries.
Record
TrialDiff compares ClinicalTrials.gov versions 16 and 17 for NCT01441947, a Massachusetts General Hospital study titled "Cabozantinib in Women With Metastatic Hormone-Receptor-Positive Breast Cancer."
Evidence Record: evt_NCT01441947_v16_v17_1493ccf4a7cd
Canonical JSON: /events/evt_NCT01441947_v16_v17_1493ccf4a7cd.json
What Changed
The patch removed one path under the secondary-outcome module:
/protocolSection/outcomesModule/secondaryOutcomes/3.
The same amendment also updated registry housekeeping paths including status verification date, last update submit date, and last update post date. It also added an annotation section. The case is useful because the evidence object is small enough to inspect directly.
Why It Carried The Critical Triage Label
The deterministic rule secondary_outcome_any_change fired because a secondary-outcome registry path
changed. The base triage severity was high.
The from-version timing context was late_recruitment. TrialDiff's timing modifier escalated the
final triage severity from high to critical because late-stage outcome changes have a higher triage label than the
same change before recruitment.
This section describes the classifier output, not an independent validation that critical is the correct review-priority tier. v0.2.1 calibration failed, so severity is retained only as uncalibrated triage metadata.
What This Supports
This record supports a narrow factual statement: a secondary-outcome registry field changed between versions 16 and 17, the change occurred in a late-recruitment context, and the stored JSON Patch and rule hash identify why TrialDiff assigned it a deterministic triage label.
What This Does Not Support
This case does not show misconduct, sponsor intent, scientific unjustifiability, manuscript non-disclosure, or regulatory non-compliance. It also does not determine whether the registry outcome change reflects outcome switching rather than a legitimate registry correction.
The correct next question is the record's own review question: what did the secondary outcome field say before and after the amendment, and what contemporaneous public document explains the change?